Prediction of Homeostasis Model Assessment of Insulin Resistance in Japanese Subjects

Tokai J Exp Clin Med., Vol. 37, No. 4, pp. 12-16, 212 Prediction of Homeostasis Model Assessment of Insulin Resistance in Japanese Subjects Masako NEGAMI, Eiko TAKAHASHI, Hiroki OTSUKA and Kengo MORIYAMA Department of Clinical Health Science, Tokai University School of Medicine (Received June 25, 212; Accepted August 31, 212) Objective: In our previous study, multivariate analysis showed that body mass index (BMI), triglycerides (TG) levels, and systolic blood pressure (SBP), in addition to fasting plasma glucose (FPG) levels, significantly correlated with homeostasis model assessment of insulin resistance (HOMA-IR). Our aim was to develop a predictive tool for HOMA-IR using tests of the specific health examinations. Methods: We enrolled 7,248 Japanese adults (3,793 men, and 3,455 women) in this cross-sectional study. A multiple regression model for predicting HOMA-IR was created using laboratory tests and lifestyle habits in the specific health examination. Results: HOMA-IR prediction index was developed using routinely measured parameters (BMI; levels of FPG, TG, and high-density lipoprotein cholesterol; BP; exercise; and physical activity), and did not require measurement of immunoreactive insulin levels. The prediction accuracy of the model was considered good, as indicated by R 2 values (men,.465; women,.45). Conclusion: The predictive tool for HOMA-IR using specific health examination tests allows healthcare professionals to estimate an individual s overall risk of metabolic syndrome (MetS). We propose that this tool can be used as an aid in health guidance for MetS. Key words: insulin resistance, HOMA-IR, prediction, health guidance, metabolic syndrome INTRODUCTION Resistance to insulin develops as a result of unhealthy lifestyles, and obesity is a key event in the origin and progression of the metabolic syndrome (MetS) [1-3]. For primary prevention of the Mets, it is important to identify subjects at risk of MetS, namely, those with insulin resistance. Homeostasis model assessment of insulin resistance (HOMA-IR) is a useful model for assessing insulin resistance by a single measurement of fasting plasma glucose (FPG) and immunoreactive insulin (IRI) levels [4]. HOMA-IR has been validated using the euglycemic hyperinsulinemic clamp method [5, 6], which is an expensive and invasive gold-standard method. We have previously identified anthropometric and metabolic parameters other than FPG in a non-diabetic Japanese population that showed a significant correlation with HOMA-IR on multivariate analyses [7]. Body mass index (BMI), triglycerides (TG) levels, and systolic blood pressure (SBP) significantly correlated with insulin resistance in men and women. High-density lipoprotein cholesterol (HDL-C) levels were found to be correlated with insulin resistance by multiple logistic regression analysis in men. Insulin measurement is still expensive and difficult to perform in some healthcare settings. In Japan, basic health examinations were performed in April 28 with specific health examination and guidance aimed for subjects, aged between 4 and 74, who are covered by health insurance. This examination and guidance includes advice to prevent MetS and detect lifestyle-related diseases at an early stage; however, this examination does not include measurement of insulin levels. Therefore, development of a predictive tool for HOMA-IR that estimates insulin resistance using routinely measured parameters will be clinically useful. The aim of the present study was to propose a predictive tool for determining insulin resistance to efficiently identify subjects at a risk of MetS. A multiple regression model for predicting HOMA-IR was created using tests of specific health examinations. METHOD Subjects Between April 27 and March 212, 9,15 people (men, 5,28 and women, 4,122) underwent a first annual health examination at the Health Evaluation and Promotion Center at Tokai University Hachioji Hospital. Of these, 7,248 adults (men, 3,793 and women, 3,455) were enrolled in this study after excluding subjects with an FPG level 126 mg/dl; those on medication for diabetes, hypertension, and dyslipidemia; or those with a history of coronary artery disease, cerebrovascular disease and chronic renal failure. The medical history of subjects was assessed using self-administered questionnaires and interviews by nurses. Verbal consent was obtained from the subjects for using their health records for analysis. Our study was a cross-sectional study approved by the Ethics Committee of Tokai University School of Medicine; further, our study conforms to the principles outlined in the Declaration of Helsinki. Eiko TAKAHASHI, Health Evaluation and Promotion Center, Tokai University Hachioji Hospital, 1838 Ishikawa-machi, Hachioji, Tokyo 192-32, Japan Tel: +81-42-639-1111 Fax: +81-42-639-1178 E-mail: etaka@tokai.ac.jp 12

Measurements Anthropometric measurements and blood sampling were performed after overnight fasting. Blood pressure was measured using an automatic blood pressure monitor (TM-2655P; A&D Co., Ltd., Tokyo, Japan) on the right upper arm in a sitting position. BMI was calculated by dividing weight (kg) by height squared (m 2 ). Waist circumference (WC) was measured at the level of the umbilicus while standing, and during slight expiration. Serum levels of low-density lipoprotein cholesterol (LDL-C), HDL-C, and TG were measured by visible spectrophotometry (Determiner L LDL-C, Determiner L HDL-C, Determiner L TG II; Kyowa Medex Co., Ltd., Tokyo, Japan). Levels of aspartate aminotransferase (AST), alanine aminotransferase (ALT), and γ-glutamyl transpeptidase (γ-gt) were measured using an automatic analyzer (JCA-BM225; JEOL, Tokyo, Japan). All measurements were included in the routine health examinations. Assessment of insulin resistance Fasting serum IRI levels were measured by fluorescence-enzyme immunoassay (ST AIA-PACK IRI; Toso Co., Ltd., Tokyo, Japan). The intra- and interassay coefficients of variation were 1.4-2.3% and 2.6-4.6%, respectively, and cross-reactivity with proinsulin molecules was 2.%. HOMA-IR was calculated as follows: fasting plasma glucose level IRI level (µu/ml)/45 [4]. Statistical analysis To design a dummy-variable multiple regression model, a series of binary (ie, dummy) variables was created that identify whether or not an observation belongs to a specific category. Binary variables were coded as 1 or. BMI was classified into 5 categories: < 22 kg/m 2, 22 to < 24 kg/m 2, 24 to < 26 kg/m 2, 26 to < 28 kg/m 2, or 28 kg/m 2. FPG levels were classified into 3 categories: < 1 mg/dl, 1 to < 11 mg/dl, or 11 mg/dl. TG levels were classified into 3 categories: < 15 mg/dl, 15 to < 3 mg/dl, or 3 mg/dl. HDL-C levels were classified into 3 categories: 5 mg/dl, 4 to < 5 mg/dl, or < 4 mg/dl. SBP was classified into 2 categories: < 13 mmhg or 13 mmhg. Diastolic blood pressure (DBP) was classified into 2 categories: < 85 mmhg or 85 mmhg. ALT levels were classified into 3 categories: < 31 U/L, 31 to < 51 U/L, or 51 U/L. If there were 4 categories, 3 dummy variables were created. The lowest category was used as a reference for BMI, FPG levels, TG levels, SBP, DBP and ALT levels, and the highest category was used as a reference for HDL-C levels. Multiple linear regression analysis was performed to find significant determinants of HOMA-IR, including BMI, FPG levels, TG levels, HDL-C levels, SBP, DBP, ALT levels, exercise ( 3 min a time, 2 times/week), physical activity ( 1 h/day), fast walker, fast eater, eating supper within 2 h before bedtime, snacking after supper, and skipping breakfast. Variables were selected in a stepwise procedure (p <.1). Data are presented as the mean ± standard deviation (SD). SAS Software version 9.2 (SAS Institute Inc., Cary, NC, USA) was used for the statistical analyses. All p values were two-tailed, and p <.5 was considered significant. RESULTS The clinical characteristics of the subjects are shown in Table 1. The mean values of HOMA-IR were 1.55 for men and 1.28 for women. On the basis of our previous study [7], BMI was used as a variable related to insulin resistance as an index for obesity. Table 2 shows the results of multivariate analyses including BMI, FPG levels, TG levels, HDL-C levels, SBP, DBP, ALT levels, exercise, physical activity, fast walker, fast eater, eating supper within 2 h before bedtime, snacking after supper, and skipping breakfast as independent variables. Stepwise multiple linear regression analysis was performed to find significant determinants for HOMA-IR. Although BMI, FPG, TG, HDL-C, ALT, SBP, exercise, and physical activity were selected, DBP was included only in women. The regression coefficients for each variable are shown in Table 2, where BMI, < 22 kg/m 2 ; FPG level, < 1 mg/dl; TG level, < 15 mg/dl; SBP < 13 mmhg, DBP < 85 mmhg; ALT level, < 31 U/L; and HDL-C level, 5 mg/dl were used as references. If the subject was a man with BMI of 25 kg/m 2, FPG level of 15 mg/dl, ALT of 4 U/L, HDL-C level of 55 mg/dl, TG level of 125 mg/dl, SBP of 14 mmhg, and no exercise and no physical activity, the predicted probability is calculated as follows:.462 +.356 +.259 + + +.137 + + +.815 = 2.331. R 2 (coefficient of determination) showed a good predictive efficacy (men,.465; women,.45). DISCUSSION We developed a method for predicting HOMA-IR using BMI, FPG levels, TG levels, HDL-C levels, ALT levels, SBP, exercise, and physical activity (and DBP for women only). One of the main strengths of our model is that it enables us to predict insulin resistance using routinely measured parameters, without having to measure IRI. On the basis of our previous study, BMI, levels of FPG, TG, and HDL-C, and SBP were used as variables related to insulin resistance for men, and BMI, levels of FPG, and TG, and SBP were used for women by multiple linear regression analysis including 4 components of MetS (obesity, glucose, blood pressure, and lipid) as independent variables [7]. The selected components were similar to the existing MetS criteria, except for the measure of obesity, for which BMI had a stronger association than WC with insulin resistance. Therefore, BMI was used as an index of obesity in this study too. In addition, it is noteworthy that MetS risk factors were stratified using dummy variables, but not contentious variables in the HOMA-IR prediction index, and that ALT levels, exercise and physical activity were selected in addition to the components of MetS for the index. Adipose tissue-derived secretory proteins are collectively named adipocytokines. Obesity and mainly visceral fat accumulation impair adipocyte function and adipocytokine secretion [8]. Impaired adipocytokine secretion promotes hepatic steatosis and development of nonalcoholic steatohepatitis [9]. Adiponectin directly regulates glucose metabolism and insulin 13

Table 1 Background characteristics of study subjects Age (years) Height (cm) BMI (kg/m 2 ) WC (cm) SBP DBP FPG F-IRI (µu/ml) HOMA-IR LDL-C HDL-C TG AST (U/L) ALT (U/L) γ-gt (U/L) Exercise ( 3 min at a time, 2 times/week) Physical activity ( 1 h/day) Fast walker Fast eater Eating supper within 2 h before bedtime Snacking after supper Skipping breakfast Men (n = 3,793) Women (n = 3,455) p * 47.6 ± 11.2 47.3 ± 11..253 17.6 ± 6.1 157.4 ± 5.6 <.1 23.5 ± 3.1 21.4 ± 3.1 <.1 84. ± 8.5 77.3 ± 8.8 <.1 117.6 ± 16.2 111.4 ± 16.7 <.1 75.3 ± 12.2 68.7 ± 11.4 <.1 99.1 ± 8.4 93.9 ± 8.1 <.1 6.23 ± 4.8 5.43 ± 3.45 <.1 1.55 ± 1.9 1.28 ±.9 <.1 123.3 ± 31.1 116. ± 32.5 <.1 57.9 ± 14.2 72.9 ± 16.5 <.1 121.9 ± 97. 76.9 ± 45. <.1 22.7 ± 1.4 19.7 ± 1.7 <.1 27.1 ± 19.6 17. ± 14. <.1 46.9 ± 58.9 22.2 ± 27.2 <.1 22.% 18.% <.1 41.% 42.6%.181 59.3% 47.7% <.1 47.4% 33.9% <.1 43.3% 18.2% <.1 15.2% 18.3% <.1 2.6% 12.5% <.1 Data are means ± standard deviation (SD). * : t-test or chi-square test BMI, body mass index; WC, waist circumference; SBP, systolic blood pressure; DBP, diastolic blood pressure; FPG, fasting plasma glucose; FIRI, fasting immunoreactive insulin; HOMA-IR, homeostasis model assessment of insulin resistance; LDL-C cholesterol, low-density lipoprotein cholesterol; HDL-C cholesterol, high-density lipoprotein cholesterol; TG, triglyceride; AST, aspartate aminotransferase; ALT, alanine aminotransferase; γ -GT, γ -glutamyl transpeptidase. sensitivity in the liver and skeletal muscle [1]. Thus, accumulation of visceral fat accompanied by reduced levels of adiponectin clearly explains complication of insulin resistance and fatty liver. Therefore, ALT levels selected in the HOMA-IR prediction index would express accumulation of fat in the liver. In addition, our previous study showed that including measurement of ALT levels as an additional parameter was useful for the diagnosis of early-stage MetS [11]. Moreover, exercise and physical activity were selected in this index because insulin resistance is improved by exercise training without weight loss [12, 13]. Insulin resistance plays a crucial role in the pathophysiology of MetS [1-3], which is a complex of interrelated risk factors for cardiovascular disease. It is important to identify subjects at risk of MetS, especially in the pre-atherogenic stage, when insulin resistance contributes to the clustering of borderline metabolic risk factors. Reference values are of 2 types, the most common one referred to as healthassociated, which is derived from a reference sample of individuals who are in good health. The other type is referred to as decision-based, and defines specific medical decision limits that are used by clinicians to diagnose or manage patients. We have established the reference interval for HOMA-IR as between.4 and 2.4, and proposed that HOMA-IR 2.5 be considered a reasonable indicator of insulin resistance in a Japanese population [14]. Further, we reported that the optimal cut-off value for HOMA-IR for diagnosing MetS in non-diabetic Japanese subjects is 1.7 [15]. Our model can be particularly useful when subjects have multiple mild abnormalities that act synergistically to increase insulin resistance. Another advantage of our model is that it enables us to assess an individual's overall metabolic status by a single indicator, insulin resistance. Our study had some limitations. First, our model has not been externally validated, and whether our model can be applied to other groups that show considerably different degrees of insulin resistance remains to be clarified. Another limitation of this study is that HOMA-IR was used as an index of insulin resistance, because it sometimes fails to show a close relationship with whole body insulin resistance assessed using the euglycemic clamp method, especially in subjects with high FPG levels [5, 16]. Because of the cross-sectional nature of the present study, a prospective study is required to show the risk of MetS in future. Finally, our model includes parameters that require fasting, such as FPG and TG. If a model could be created using parameters without using fasting values, it might be of greater utility not only in hospitals but also in community or industrial health care settings. In conclusion, HOMA-IR prediction index was developed using routinely measured metabolic parameters that allows healthcare professionals to estimate an individual's overall MetS risk. We propose that this 14

Table 2 Results of multivariate linear regression analyses (A) Men Coefficient SE F-value p BMI (kg/m 2 ) FPG ALT (U/L) HDL-C TG SBP < 22 22 to < 24 24 to < 26 26 to < 28 28 < 1 1 to < 11 11 < 31 31 to < 51 51 5 4 to < 5 < 4 < 15 15 to < 3 3 < 13 13 Exercise < 2 times/week 3 min at a time, 2 times/week Physical activity < 1 h/day 1 h/day.219.462.764 1.491.356.839.259.84.246.294.129.285.34.39.5.6.3.44.37.53.33.56.35.76 4.81 14.65 236.18 622.29 142.79 364.44 49.8 228.83 56.66 27.56 13.4 14.5.3.2.137.33 17.17 -.191.32 35.75 -.75.27 7.75.54 Intercept.815.31 68.82 SE, standard error; BMI, body mass index; FPG, fasting plasma glucose; ALT, alanine aminotransferase; HDL-C cholesterol, high-density lipoprotein cholesterol; TG, triglyceride; SBP, systolic blood pressure. (B) Women Coefficient SE F-value p BMI (kg/m 2 ) FPG ALT (U/L) HDL-C TG SBP DBP < 22 22 to < 24 24 to < 26 26 to < 28 28 < 1 1 to < 11 11 < 31 31 to < 51 51 5 4 to < 5 < 4 < 15 15 to < 3 3 < 13 13 < 85 85 Exercise < 2 times/week 3 min at a time, 2 times/week Physical activity < 1 h/day 1 h/day.241.439.83 1.613.435.912.455.473.236.329.32 1.332.32.42.64.68.33.58.57.14.54.125.55.225 58.12 111.39 168.38 563.81 171.1 243.61 63.7 2.59 19.19 6.94 3.7 35.7.84.84.41 4.8.434.183.49 14.1.2 -.15.31 23.3 -.42.24 3.6.81 Intercept.946.2 2259.46 SE, standard error; BMI, body mass index; TG, triglyceride; FPG, fasting plasma glucose; ALT, alanine aminotransferase; HDL-C cholesterol, high-density lipoprotein cholesterol; SBP, systolic blood pressure; DBP, diastolic blood pressure. model can be used as an aid in health guidance for MetS. ACKNOWLEDGEMENTS The authors' responsibilities were as follows; ET designed the research, KM and ET conducted the research, MN and ET analyzed data, MN wrote the 15

manuscript, ET had primary responsibility for the final content of the manuscript, and all authors read and approved the final manuscript. The authors declare no conflicts of interest. REFERENCES 1) Reaven GM. Role of insulin resistance in human disease. Diabetes 1988; 37: 1595-167. 2) DeFronzo RA, Ferrannini E. Insulin resistance. A multifaceted syndrome responsible for NIDDM, obesity, hypertension, dyslipidemia, and atherosclerotic cardiovascular disease. Diabetes Care. 1991; 14: 173-194. 3) Haffner SM, Valdez RA, Hazuda HP, Mitchell BD, Morales PA, Stern MP. Prospective analysis of the insulin-resistance syndrome (syndrome X). Diabetes 1992; 41: 715-722. 4) Matthews DR, Hosker JP, Rudenski AS, Naylor BA, Treacher DF, Turner RC. Homeostasis model assessment: insulin resistance and beta-cell function from fasting plasma glucose and insulin concentrations in man. Diabetologia 1985; 28: 412-419. 5) Wallace TM, Levy JC, Matthews DR. Use and abuse of HOMA modeling. Diabetes Care 24; 27: 1487-1495. 6) Bonora E, Targher G, Alberiche M. Homeostasis model assessment closely mirrors the glucose clamp technique in the assessment of insulin sensitivity: studies in subjects with various degrees of glucose tolerance and insulin sensitivity. Diabetes Care 2; 23: 57-63. 7) Yamada C, Moriyama K, Takahashi E. Association between insulin resistance and metabolic syndrome risk factors in Japanese. J Diabetes Invest 212; 3: 185-19. DOI: 1.1111/j. 24-1124. 211. 162. x 8) Diez JJ, Iglesias P. The role of the novel adipocyte-derived hormone adiponectine in human disease. Eur J Endocrinol 23; 148: 293-3. 9) Schaffler A. Scholmerich J, Buchler C. Mechanism of disease: adipocytokines and visceral adipose tissue-emerging role inch nonalcoholic fatty liver disease. Nat Clin Pract Gastroenterol Hepatol 25; 2: 273-28. 1) Yamauchi T, Kamon J, Minokoshi Y, Ito Y, Waki H, Uchida S, et al. Adiponectine stimulates glucose utilization and fatty-acid oxidation by activating AMP-activated protein kinase. Nat Med 22; 8: 1288-1295. 11) Yamada C, Inabe F, Mitsuhashi T, Hiratsuka N, Araida N, Takahashi E. Liver Enzymes as Indicators for the Diagnosis of Metabolic Syndrome and Its Development. Ningen Dock 21; 26: 29-36. 12) Duncan GE, Perri MG, Theriaque DW, Hutson AD, Eckel RH, Stacpoole PW. Exercise training, without weight loss, increases insulin sensitivity and postheparin plasma lipase activity in previously sedentary adults. Diabetes Care 23; 26: 557-562. 13) Houmard JA, Tanner CJ, Slentz CA, Duscha BD, MaCartney JS, Kraus WE. Effect of the volume and intensity of exercise training on insulin sensitivity. J Appl Physiol 24; 96: 11-16. 14) Yamada C, Mitsuhashi T, Hiratsuka N, Inabe F, Araida N, Takahashi E. Optimal reference interval for homeostasis model assessment of insulin resistance (HOMA-IR) in a Japanese population. J Diabetes Invest 211; 2: 373-376. DOI: 1.1111/j. 24-1124. 211. 113. x 15) Yamada C, Moriyama K, Takahashi E. Optimal cut-off point for homeostasis model assessment of insulin resistance to discriminate metabolic syndrome in non-diabetic Japanese subjects. J Diabetes Invest 212; 3: 384-387. DOI: 1. 1111/j. 24-1124. 212. 194. x 16) Tajiri Y, Sato S, Kato T, Nakayama H, Yamada K. Surrogate index for insulin sensitivity composed of factors not using glucose and insulin in Japanese patients with diabetes. J Diabetes Invest 211; 2: 14-147. DOI: 1. 1111/j. 24-1124. 21. 76. x 16